v2.2.3 - 2026-10-09

Added

• Added FragmentedResidueError for when multiple residues share the
  same ResidueId.
• Added OptionalException base class to define the behaviour on specific runtime
  errors that may be ignored or handled by the user.

Fixed

• MoleculeStandardizer now properly assigns each atom's hybridization.
• The HBond acceptor SMARTS patterns now exclude proline backbone nitrogens, and
  the implicit and explicit patterns are harmonized: the peptide nitrogen
  exclusion prefers degree over valence.
• Implicit HBond interactions no longer raise a KeyError for atoms with a
  hybridization other than SP, SP2 or SP3, and instead treats the interaction
  as invalid.
• Using split_molecule could swap atoms between the two output molecules if the
  components residues were ordered differently from the original molecule, typically
  when water molecules in multiple chains are present.
• The residue fragmentation that occurs when creating a Molecule no longer assigns a
  wrong hydrogen count to atoms at the cut peptide bonds when both residues are the same
  e.g. when splitting a di-Alanine.
• Fixed a bug in the "queue" parallel strategy where the pickling/unpickling of the
  RDKit molecules would silently redo the residue splitting without preserving the
  original use_segid flag.
• Fixed an issue in the "queue" parallel strategy where exceptions raised during
  trajectory iteration would silently crash the background task and hang the process
  indefinitely.
• 5 and 6-membered rings fused to a smaller ring are no longer excluded.
• Switched to SSSR ring-perception for more reliable ring size detection.

Changed

• The use_segid option is now propagated through Residue,
  split_mol_by_residues, the template engines and split_molecule, so
  residues created from these code paths keep the correct chain identifier.